Skip to main content
. Author manuscript; available in PMC: 2021 Jan 1.
Published in final edited form as: Neuroinformatics. 2020 Jan;18(1):59–70. doi: 10.1007/s12021-019-09426-x

Table 1.

Pearson’s correlation of different techniques at 20% missing data with the gold standard under MCAR and MNAR assumptions. Degrees of freedom = 12220.

missingness Correlation (r) t-statistics confident interval p-value
Soft Impute Random 0.92 254.2 0.914 – 0.919 < 0.0001
Not random 0.66 98.4 0.650 – 0.670 < 0.0001
Mean Random 0.87 199.1 0.870 – 0.878 < 0.0001
Not random 0.88 205.3 0.876 – 0.884 < 0.0001
K-NN Random 0.92 255.1 0.914 – 0.920 < 0.0001
Not random 0.92 256.4 0.915 – 0.920 < 0.0001
RF Random 0.93 275.1 0.925 – 0.930 < 0.0001
Not random 0.93 275.4 0.925 – 0.930 < 0.0001